Artificial intelligence (AI) and machine learning are increasingly applied across urology, with particular promise for the management of neurogenic lower urinary tract dysfunction (NLUTD) in individuals with spinal cord injury (SCI). There is also particular promise in the management of NLUTD, a complex, lifelong condition affecting individuals with SCI and also occurring in the context of multiple sclerosis, Parkinson's disease, and other neurological disorders. This article provides a comprehensive analysis of AI's current and emerging role in neuro-urology, with dedicated focus on the specific challenges and opportunities presented by NLUTD and SCI. We review the epidemiological and clinical burden of neurogenic bladder dysfunction. We further examine AI applications in uro-oncology, functional urology, digital pathology, and robotic surgery, integrating these with the specialized perspective of neuro-urology. The particular vulnerability of individuals with SCI to algorithmic bias, access inequity, and assistive technology dependency is critically discussed. We identified five key domains where AI may transform care for patients with NLUTD: automated urodynamic interpretation including detection of detrusor overactivity and detrusor-sphincter dyssynergia, predictive risk stratification for renal deterioration, closed-loop neuromodulation, remote digital monitoring, and AI-assisted rehabilitation support. Alongside the substantial opportunities, we articulate the ethical, moral, and societal responsibilities that accompany AI integration in this population, emphasizing that patients with SCI and NLUTD deserve not merely inclusion in AI development but active partnership as co-designers of the technologies that will shape their lives.
Abstract Background Patients increasingly use large language models (LLMs) to obtain medical information, but the quality of LLM-generated information on complex spinal infections such as spondylodiscitis remains uncertain. Existing evaluations in spine surgery have mainly addressed degenerative conditions or surgical procedures, and disease-specific data for spondylodiscitis are limited. Objective This preliminary study evaluated spine surgeons’ ratings of single-turn LLM responses to 10 author-curated frequently asked questions (FAQs) about spondylodiscitis and compared answer sets generated from GPT-4, GPT-4o, and Google Gemini web interfaces under the authors’ implemented prompting conditions. Methods A pool of patient-oriented questions was generated through a chronological workflow including publicly available FAQ sources, PubMed-informed terminology review, Google Trends topic checking, and LLM-generated candidate questions. Duplicate and semantically overlapping questions were removed, the remaining questions were grouped into thematic categories, and 10 final FAQ-style prompts were synthesized. Each prompt was submitted once to the publicly accessible web interfaces of GPT-4, GPT-4o, and Google Gemini. The study was interpreted as an expert evaluation of the resulting answer sets. Seven blinded board-certified spine surgeons rated the responses using a 4-level rating system ranging from excellent to unsatisfactory and additionally assessed comprehensiveness, clarity, empathy, and appropriateness of length. Descriptive statistics and nonparametric comparisons were performed. Interrater reliability was assessed using the intraclass correlation coefficient. Results Across all responses, 38.6% (81/210) were rated as excellent, 39% (82/210) as satisfactory with minimal clarification needed, 16.7% (35/210) as satisfactory with moderate clarification needed, and 5.7% (12/210) as unsatisfactory. The most common reason for necessary clarification was insufficient information (58/141, 41.1%), followed by language-related issues (21/141, 14.9%) and overly detailed responses (18/141, 12.8%). The complication-related question received the highest mean rating (3.4/5), whereas treatment- and prognosis-related questions received lower ratings (2.7/5 and 2.9/5). Median overall ratings did not differ significantly among the 3 evaluated LLMs. Spine surgeons reported a generally positive attitude toward artificial intelligence–supported patient information but expressed remaining uncertainty regarding reliability and direct patient-physician communication. Conclusions In this preliminary expert evaluation, selected publicly accessible LLM web interfaces generated mostly satisfactory responses to author-curated spondylodiscitis FAQs. The findings reflect outputs produced at the time of access under the authors’ implemented prompting conditions. LLMs may support patient education only with clinician oversight. Future research should explore advanced, domain-specific models to further improve the quality of communication between clinicians and patients.
Clinical data are split across incomplete cohorts with partially overlapping variables. Causal discovery under missing data and multi-source integration have developed separately, and tabular foundation models cannot take a causal graph without retraining. We make two contributions. First, causal graph fusion learns a partial ancestral graph on the shared variables, fuses it into each cohort's graph under temporal background knowledge and lets Markov blankets guide re-imputation, so missing data are handled inside discovery. Second, causal attention steering compiles the learned graph into the attention of a frozen tabular foundation model at inference. Against known ground truth, fusion dominates pooled and per-cohort baselines in every missingness and heterogeneity condition tested (p<0.001), and steering improves discrimination under distribution shift (p<10^-5), a gain that a random graph does not produce. In two spinal cord injury cohorts, fusion orients age, albumin and hemoglobin as ancestors of length of stay, a set no single cohort orients in full, and yields compact, temporally ordered predictor sets that match or exceed automatic feature selection on four of five outcomes.
This systematic review and meta-analysis aimed to evaluate the safety and efficacy of permanent peritoneal catheters (PPC) and peritoneal ports (PP) in patients with refractory ascites of malignant and non-malignant etiology, irrespective of etiology, and to compare these devices to large-volume paracentesis (LVP). Twenty-nine studies were included, with the majority being case series. Meta-analysis found no significant difference in peritonitis risk between PPC/PP and LVP. Cellulitis, occlusion, and ascites leakage were variably reported across studies. Average catheter patency was approximately 72 days. Quality of life improvements were reported in only a subset of studies, with considerable heterogeneity in assessment tools and limited standardization of definitions and prophylaxis protocols. Despite substantial heterogeneity and risk of bias across included studies, no significant differences in safety were observed between PPC/PP and LVP. Available evidence suggests these devices may offer practical advantages in symptom control and could reduce the need for repeated paracentesis in selected patients. However, further randomized studies are needed to standardize protocols and better guide clinical implementation. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. This protocol was register with PROSPERO (CRD42024622692).
An observational study with the Swiss Paraplegic Centre, Nottwil, aimed to tailor robotic systems to therapists’ preferences by quantifying therapist-patient interactions, as understanding the therapists’ adaptation strategies is essential for effective customization. A total of 2297 interactions were recorded in 198 exercises with 16 patient-therapist pairs, either directly or through a robotic device with enhanced analysis tools. Derived correlation maps quantified three preferred interaction strategies: Observation of compensatory movement and posture issues caused direct tactile reaction, issues with robotic settings led to robotic support adaptations, and robotic support adaptations preceded additional support adaptations without intermediate patient observation. Strategies varied by exercise type, with direct tactile reactions dominating reach-goal exercises and robotic adjustments dominating in nominal path exercises. We conclude that a limited set of robotic feedback and adaptation options fails to meet therapeutic needs, given the wide variability in individual rehabilitation strategies. In contrast, offering an exhaustive range of options overwhelms therapists, often causing them to favor direct patient interaction over engaging with the robotic system. Therefore, an optimal robotic system must incorporate methods that understand the therapist’s strategic intent and dynamically adapt the presented information and adjustable parameters to the specific needs and preferences of the operating therapist.